Rifampin modulation of xeno‐ and endobiotic conjugating enzyme mRNA expression and associated microRNAs in human hepatocytes

Abstract Rifampin is a pleiotropic inducer of multiple drug metabolizing enzymes and transporters. This work utilized a global approach to evaluate rifampin effects on conjugating enzyme gene expression with relevance to human xeno‐ and endo‐biotic metabolism. Primary human hepatocytes from 7 subjects were treated with rifampin (10 μmol/L, 24 hours). Standard methods for RNA‐seq library construction, EZBead preparation, and NextGen sequencing were used to measure UDP‐glucuronosyl transferase UGT, sulfonyltransferase SULT, N acetyltransferase NAT, and glutathione‐S‐transferase GST mRNA expression compared to vehicle control (0.01% MeOH). Rifampin‐induced (>1.25‐fold) mRNA expression of 13 clinically important phase II drug metabolizing genes and repressed (>1.25‐fold) the expression of 3 genes (P < .05). Rifampin‐induced miRNA expression changes correlated with mRNA changes and miRNAs were identified that may modulate conjugating enzyme expression. NAT2 gene expression was most strongly repressed (1.3‐fold) by rifampin while UGT1A4 and UGT1A1 genes were most strongly induced (7.9‐ and 4.8‐fold, respectively). Physiologically based pharmacokinetic modeling (PBPK) was used to simulate the clinical consequences of rifampin induction of CYP3A4‐ and UGT1A4‐mediated midazolam metabolism. Simulations evaluating isolated UGT1A4 induction predicted increased midazolam N‐glucuronide exposure (~4‐fold) with minimal reductions in parent midazolam exposure (~10%). Simulations accounting for simultaneous induction of both CYP3A4 and UGT1A4 predicted a ~10‐fold decrease in parent midazolam exposure with only a ~2‐fold decrease in midazolam N‐glucuronide metabolite exposure. These data reveal differential effects of rifampin on the human conjugating enzyme transcriptome and potential associations with miRNAs that form the basis for future mechanistic studies to elucidate the interplay of conjugating enzyme regulatory elements.


| INTRODUCTION
Rifampin induction of cytochrome P450 is an extensively studied drug-drug interaction mechanism resulting in a substantial list of clinically important interactions that can lead to reduced drug efficacy or increased toxicity. 1,2 In contrast, relatively less is known about rifampin induction of human conjugating enzymes including uridine diphosphate glucuronosyltransferases (UGTs), sulfotransferases (SULTs), N-acetyltransferases (NATs), thiopurine S-methyltransferase (TPMT) and glutathione S-transferases (GSTs). 3 Rifampin is widely recognized as a pleiotropic but specific inducer of drug metabolizing enzymes and transporters with effects mediated mainly through activation of pregnane X receptor (PXR). 4 The genes regulated by PXR include those encoding for human conjugating enzyme families (UGTs, SULTs, NATs, and GSTs). Previous studies demonstrated rifampin induction of miRNAs and association with repression of P450 genes, suggesting the possibility of additional epigenetic mechanisms underlying rifampin drug-drug interactions. 5,6 Epigenetic modulation of conjugating enzymes by miRNAs has also been demonstrated. 7 The UGT superfamily of conjugating enzymes contains 5 subfamilies (UGT1, UGT2A, UGT2B, UGT3, and UGT8). Three of these subfamilies (UGT1, UGT2A, and UGT2B) prominently contribute to the metabolism of drugs, dietary substances, toxicants, and endogenous substrates with broad and overlapping substrate specificities. These 3 subfamilies are encoded by 10 genes to generate 19 isoforms in humans. 11 The UGT1A family shares a single chromosomal locus (band 2q37) with the 9 different functional isoforms being generated via splicing of shared exons 2-5 to an isoform-specific exon 1. Similarly, the UGT2A subfamily members share exons 2-6 with an isoform-specific exon 1. Conversely, the UGT2B family is composed of 7 functional enzymes encoded by individual genes. Each UGT possesses a unique 5 0 -upstream promoter region that controls its transcription as well as more distant enhancer regions containing transcription factor-binding sites that further control constitutive and inducible UGT expression. A wide variety of tissue-specific and ligand-activated transcription factors modulate the induction of UGT genes including PXR. 12 In addition, epigenetic UGT regulation by miRNAs has recently been identified as another factor that modulates UGT expression and response to environmental exposures. [7][8][9][10]13,14 Taken together, evaluating the influence of rifampin on UGT mRNA expression and association with miRNA changes may help to unravel the complex regulatory network governing UGT expression and activity.
The cytosolic SULT family of enzymes contribute to the metabolism of several exogenous and endogenous substrates, including the clinically used drugs acetaminophen, minoxidil, and ethinyl estradiol.
The SULT family is comprised of 13 members within 3 families (SULT1, SULT2, and SULT4). SULT activity varies widely among individuals due in part to genetic polymorphisms and susceptibility to induction via nuclear receptor activation. [15][16][17] For women taking ethinyl estradiol, rifampin induction of SULTs may cause therapeutic failure of the oral contraceptive drug. 18 Despite the clinical importance of SULT-mediated xenobiotic metabolism, data describing mechanisms regulating SULT induction are rather sparse.
NATs, another family of conjugating enzymes, contribute to human xenobiotic and endogenous substrate metabolism. Two NATs, NAT1 and NAT2, are thought to be of primary importance to drug metabolism. Polymorphisms exist in both NAT1 and NAT2 genes with well-established functional consequences in phenotypic slow acetylators. For example, slow acetylators are more susceptible to drug-induced toxicities from hydralazine and isoniazid. Isoniazid and rifampin are also commonly coadministered for the treatment of latent tuberculosis, raising the potential for drug-drug interactions.
Slow acetylators are also more prone to developing certain cancers. 19 As a result, NAT modulation via small molecules and miRNAs has become a target of drug and biomarker development. 20,21 Considered together, understanding the rifampin-induced changes in NAT expression and associated miRNAs may be of therapeutic and diagnostic value.
TPMT is the primary enzyme responsible for human metabolism of thiopurine drugs including azathioprine, thioguanine, and 6-mercaptopurine. Genetic polymorphisms in TMPT can result in reduced enzyme activity leading to increased drug concentration and toxicities in certain patients. As a result, pharmacogenetics screening for TPMT deficiency is recommended prior to initiating thiopurine drug therapy. A previous report demonstrated no change in TPMT mRNA expression in human hepatocytes treated with rifampin 3 but the potential influence of miRNAs has not been previously explored.
Human GSTs are a family of cytosolic enzymes that catalyze the transfer of the sulfhydryl group of glutathione to a large variety of electrophiles, including drug molecules such as busulfan and ethacrynic acid and reactive CYP450 metabolites such as N-acetyl-p-benzoquinone imine (NAPQI). GST induction by drug molecules and dietary flavonoids has been previously reported 22,23 but the potential relationship with miRNA expression changes has not been evaluated.
The first aim of this report was to describe the effects of rifampin treatment on the regulation of hepatic conjugating enzyme mRNA expression and the relationships with regulation of miRNA expression in primary human hepatocytes. The second aim was to further assess the impact of rifampin modulation of UGT mRNA expression in human renal proximal tubule cells to evaluate the potential for tissue-specific changes in enzyme regulation. Finally, based upon the in vitro and in silico study results, rifampin induction of UGT1A4-mediated metabolism was selected for further evaluation via physiologically based pharmacokinetic (PBPK) modeling and simulation. The overarching goal of this work was to globally evaluate rifampin's effects on conjugating enzyme gene expression with relevance to human xeno-and endobiotic metabolism.  The remaining miRNA bioinformatics analyses mirrored that described in a previous analysis of this data set for evaluation of transport protein changes. 5 2.3 | Bioinformatic analysis of the RNA-seq data RNA-seq library construction, EZBead preparation, and NextGen sequencing were performed using standard methods as described previously 6 and used to measure UGT, SULT, NAT, TPMT, and GST mRNAs and compared to vehicle control (0.1% methanol).
UGT1A genes were identified and quantified by unique exons 1 as exons 2-5 are shared across this gene subfamily. The RNA-Seq data analysis included quality assessment and sequence alignment prior to differential gene expression analysis as described previously. 6 In brief, SOLiD Instrument Control Software and Experiment Tracking Software were used for read quality recalibration.
Each sequence was scanned for low-quality reads and any read length of less than 35 bases was discarded to effectively eliminate low-quality reads while retaining high-quality regions. BFAST was used as the primary sequence alignment algorithm employing a TopHat-like strategy to align sequencing reads that crossed splicing junctions. Sequence reads were aligned to a filtering index to exclude sequences that were not of interest (eg, repeats and ribosomal RNA). Analyses were restricted to uniquely aligned sequences with 2 or less mismatches. Differentially expressed genes were identified using edgeR following exclusion of genes with less than 1 read per million mappable reads in more than   Table S1. The delta--delta C T method was applied to determine the relative expression of each gene for rifampin and vehicle-treated cells as previously described. 5 The fold change in gene expression is represented as the mean AE SEM of the biological replicates (n = 4).

| ChIP-seq PXR-binding site in silico analysis
The conjugating enzyme genes in this study were evaluated in silico for PXR-binding sites using a publically available ChIP-Seq database generated using HepG2 cells treated with vehicle (dimethyl sulfoxide, DMSO) or rifampin. 25 The in silico ChIP-Seq testing was conducted as described previously. 5 Promoter regions were specified as AE2 kb based upon the coordinates of each transcription start site.

| Accession numbers
Raw RNA-seq data were made publicly available through the National Center for Biotechnology Information Gene Expression Omnibus (GEO) database and can be accessed, using GEO series accession number GSE799933 (http://www.ncbi.nlm.nih.gov/ge o/query/acc.cgi?acc=GSE79933). OpenArray miRNA data were made publically available through the Indiana University Center for Computational Biology and Bioinformatics and can be accessed at http://compbio.iupui.edu/group/6/pages/rifampin.

| Physiologically based modeling and simulation
The potential clinical impact of rifampin induction of UGT1A4- within the software. Midazolam N-glucuronide model development was accomplished, using clinical data previously acquired during the control phase of a healthy volunteer (n = 12) herbal product-drug interaction study. 26 SimCYP model parameters are available in Table S2. Simulated pharmacokinetic outcomes within 30% of observed endpoints were deemed sufficiently accurate to proceed with interaction simulations. Drug-drug interactions resulting from coadministration of rifampin (600 mg/day orally for 3 days) with midazolam (5 mg orally on day 3) were simulated in 10 virtual trials 3.2 | Conjugating enzymes and CYP450 isoforms appear to be coordinately regulated Coordinate regulation of CYPs, UGTs, and transporters has been proposed as a defense mechanism providing protection against various chemical stressors. 30 Correlation analysis suggests that several conjugating enzymes are coordinately regulated in response to rifampin treatment. As expected, multiple UGT genes displayed strong positive correlations in rifampin-induced expression changes ( Table 2).  Table 2). Changes in UGT1A1 expression correlated positively with changes in CYP3A7 and CYP2B6 expression while UGT1A5 expression negatively correlated with changes in CYP3A5 and CYP2E1 (Table 3). Interestingly, significant correlations were not observed between the major CYPs (eg, CYP1A2, 2C8, 2C9, and 3A4/5) and UGTs (UGT1A1, 1A3, 1A4, 1A9, 2B7). These data provide further evidence for coordinate regulation of drug metabolizing enzymes in response to rifampin treatment.

| Physiologically based modeling and simulation
suggests that UGT induction contributes to observed rifampin-drug interactions with dual CYP3A/UGT substrates Simulated midazolam and midazolam N-glucuronide concentrationtime profiles closely approximated clinically observed disposition and pharmacokinetic outcomes (Figure 2A and B, Table 5). Simulations evaluating the impact of rifampin-induced UGT1A4 metabolism in isolation predicted markedly increased midazolam N-glucuronide exposure (~fourfold) with minimal reductions in parent midazolam exposure (~10%), consistent with midazolam clearance-mediated primarily by CYP3A4 ( Figure 2C and D,  including resveratrol, curcumin, and chrysin in Caco-2 cells, 41,42 human hepatocytes, 43 and PXR reporter assays. 44 However, rapid metabolism and minimal systemic exposure of many dietary polyphenols may limit their ability to induce hepatic UGTs in vivo. 45   hepatotoxin. 51 However, human NAT2 genetic polymorphisms that result in a slow acetylator phenotype have been strongly associated with increased risk of isoniazid hepatotoxicity. It then leads that perhaps rifampin down-regulation of NAT2 is creating a drug-induced slow acetylator phenotype that leads to increased risk of isoniazid hepatotoxicity when administered with rifampin. Rifampin-induced formation of hydrazine from isoniazid has been posited to underlie increased hydrazine plasma levels observed in patients taking rifampin and isoniazid as compared to those taking isoniazid alone. 53 Alternatively, repressed NAT2 activity leading to impaired hydrazine elimination, or a combination of both increased formation and reduced elimination, may explain the apparent increase in hydrazine exposure caused by rifampin. Further reduction in limited NAT2 activity by rifampin could potentially explain reports of increased incidence of hepatotoxicity when slow acetylators take isoniazid and rifampin. 54 The alpha-class GSTs catalyze the GSH-dependent detoxification of several alkylating chemotherapy agents and numerous environmental pollutants. 55 GST induction has also been suggested, using high-sensitivity real-time PCR 3 and likely represents another defense mechanism against xenobiotic exposure. The observed changes in GST expression measured via RNAseq are in alignment with previous reports using alternate quantification approaches. miR-133a has been associated with repression of GSTP1 mRNA and protein in lung and bladder cancer cell lines 57,58 while miR-133b has been associated with repressed GSTP1 mRNA expression in prostate cancer cell lines. 59 miR-513a-3p has also been associated with repressed GSTP1 expression in lung cancer cells. 60 An inverse correlation between PXR translational efficiency and miR-148a has also been reported. 61 Interestingly, none of these miRNAs were revealed by our correlation analysis. This may be the result of both direct and indirect mechanisms mediated via rifampin induction. However, the miRNA-mRNA pairs identified in

ACKNOWLEDG EMENTS
We thank Mary F. Paine for kindly providing midazolam and midazolam-N-glucuronide clinical pharmacokinetic data used to support PBPK model development. Observed data recovered from a healthy volunteer (n = 12) study in which participants were administered a single oral dose (5 mg) of midazolam. AUC obs , observed area under the plasma concentration-time curve (nmol/L 9 hours); AUC pred , predicted area under the plasma concentration-time curve (nmol/ L 9 hours); AUC Ind , predicted area under the plasma concentration-time curve following rifampin induction (nmol/L 9 hours); AUC ratio , rifampin treatment:control ratio; C maxobs , maximal observed plasma concentration (nmol/L); C maxpred , maximal predicted concentration (nmol/L); C maxratio , rifampin treatment:control ratio; C maxind , predicted maximal concentration following rifampin induction (nmol/L). Values denote geometric mean and 95% confidence intervals.